Evaluation of atmospheric correction using bi-temporal hyperspectral images

Atmospheric correction of hyperspectral image data is frequently a requirement for using remote sensing to understand and quantify various phenomena that take place on the Earth. This is particularly true when the analysis requires the use of spectral reflectance. Although sophisticated models exist that can be used to perform atmospheric correction, evaluating the performance of these procedures is non-trivial. In this study, two atmospheric correction programs, FLAASH (based on MODTRAN 4), and TAFKAA_6S (based on 6S), were applied to a pair of images of the same area but collected six weeks apart. The results of the two atmospheric correction procedures are analyzed based on the expected stability of pseudo-invariant features (PIFs). Although both procedures performed rather well in terms of removing atmospheric absorption features in the infrared region, the analysis identified some anomalous behaviors as well, the most important of which appears to be related to the bidirectional reflectance distribut...